A media content classification method, device, storage medium and electronic device

Abstract

The application discloses a media content classification method, comprising: obtaining a first image of a media content display area and text content and corresponding text position information in the media content display area in an original image; using a pre-trained media content classification network to extract semantic features of the text content to obtain semantic features of the text content; inputting the text position information into an embedding layer of the media content classification network to perform embedding projection on each dimension of the text position information, combining projection results of the dimensions to obtain embedding layer features of the text position information; extracting features of the first image to obtain features of the first image; connecting the semantic features, the embedding layer features of the text position information and the features of the first image together and inputting the same into an image classification layer for processing to obtain a media content classification to which the original image belongs. The application can improve the accuracy of media content classification.

Description

TECHNICAL FIELD

[0001] The present application relates to neural network technology, in particular to a media content classification method and device, a storage medium and an electronic device. BACKGROUND

[0002] With the progress of neural network technology, the technology of using neural networks for media content classification (such as classifying advertisements into advertisement categories such as commercial advertisements and public service advertisements) has been more and more widely used.

[0003] Taking advertisement classification as an example, the existing media content classification methods mainly fall into two categories. In the first category of methods, advertisement classification and illegal advertisement identification are performed using the semantics of advertisement text. In the second category of methods, illegal advertisement identification is performed according to whether the position of the advertisement board is in an illegal area. The above processing methods all have different problems:

[0004] 1) The method of using only advertisement semantics for identification cannot identify situations such as advertisement word occlusion.

[0005] 2) The method of judging according to whether the position of the advertisement board is in an illegal area is only applicable to special points and has low applicability.

[0006] As can be seen from the above, the current media content classification methods, including advertisement classification, all have various problems, resulting in low accuracy of media content classification. SUMMARY

[0007] The present application provides a media content classification method, device, storage medium and electronic device, which can improve the accuracy of media content classification.

[0008] To achieve the above-mentioned purpose, the present application adopts the following technical solutions:

[0009] A media content classification method, comprising:

[0010] obtaining a first image of a media content display area and text content and corresponding text position information in the media content display area in an original image;

[0011] processing the first image, the text content and the text position information using a pre-trained media content classification network to determine the media content classification to which the original image belongs;

[0012] wherein the processing of the first image, the text content and the text position information comprises:

[0013] performing semantic feature extraction on the text content to obtain semantic features of the text content;

[0014] input the text position information into an embedding layer of the media content classification network, perform embedding projection on each dimension of the text position information, combine the projection results of the respective dimensions to obtain an embedding feature of the text position information;

[0015] perform feature extraction on the first image to obtain a feature of the first image;

[0016] concatenate the semantic feature of the text content, the embedding feature of the text position information, and the feature of the first image together, input the concatenated result into a two-dimensional image classification layer for processing, and obtain the media content classification to which the original image belongs.

[0017] Preferably, the concatenating the semantic feature of the text content, the embedding feature of the text position information, and the feature of the first image together comprises:

[0018] concatenate the semantic feature of the text content and the embedding feature of the text position information together to obtain a first feature;

[0019] transform the feature of the first image into a two-dimensional feature, and the transformed feature has the same dimension as one dimension of the first feature;

[0020] concatenate the first feature and the transformed feature of the first image together.

[0021] Preferably, the text content and the corresponding text position information are obtained in units of text boxes, and the semantic feature extraction of the text content and the processing of the text position information by the embedding layer are both performed in units of text boxes.

[0022] Preferably, when the semantic feature of the text content and the embedding feature of the text position information are concatenated together, the semantic feature of the text content and the embedding feature of the text position information corresponding to the same text box are concatenated together.

[0023] Preferably, the text position information comprises coordinates of a specified boundary point of a text box, or coordinates of a center point of a text box and width and height information of the text box.

[0024] A media content classification apparatus comprises an obtaining unit and a classification network unit.

[0025] The obtaining unit is configured to obtain, in an original image, a first image of a media content display area and text content and corresponding text position information in the media content display area.

[0026] The classification network unit is configured to process the first image, the text content, and the text position information by using a pre-trained media content classification network to determine a media content category to which the original image belongs.

[0027] In the classification network unit, the processing of the first image, the text content, and the text position information includes:

[0028] performing semantic feature extraction on the text content to obtain a semantic feature of the text content;

[0029] inputting the text position information into an embedding layer of the media content classification network to perform embedding projection on each dimension of the text position information, combining projection results of the respective dimensions to obtain an embedding feature of the text position information;

[0030] performing feature extraction on the first image to obtain a feature of the first image;

[0031] connecting the semantic feature of the text content, the embedding feature of the text position information, and the feature of the first image together and inputting them into a two-dimensional image classification layer for processing to obtain the media content category to which the original image belongs.

[0032] Preferably, in the classification network unit, the connecting of the semantic feature of the text content, the embedding feature of the text position information, and the feature of the first image together includes:

[0033] connecting the semantic feature of the text content and the embedding feature of the text position information together to obtain a first feature;

[0034] transforming the feature of the first image into a two-dimensional feature with a same dimension as one dimension of the first feature;

[0035] connecting the first feature and the transformed feature of the first image together.

[0036] Preferably, in the obtaining unit, the text content and the corresponding text position information are obtained in a text box unit.

[0037] In the classification network unit, the semantic feature extraction on the text content and the processing of the embedding layer on the text position information are both performed in a text box unit.

[0038] Preferably, in the classification network, the semantic features of the text content and the embedding features of the text position information corresponding to the same text box are connected together when the semantic features of the text content and the embedding features of the text position information are connected together.

[0039] Preferably, the text position information is: coordinates of a specified boundary point of a text box, or coordinates of a center point of a text box and width and height information of the text box.

[0040] The application also provides a computer readable storage medium, which stores computer instructions, and the instructions are executed by a processor to implement the media content classification method.

[0041] The application also provides an electronic device, which at least includes a computer readable storage medium and further includes a processor.

[0042] The processor is configured to read the executable instructions from the computer readable storage medium and execute the instructions to implement the media content classification method.

[0043] As can be seen from the above technical solutions, the first image of the media content display area and the text content and the corresponding text position information in the media content display area can be obtained in the original image; the text content is input into the embedding layer of the pre-trained media content classification network for processing to obtain the embedding features of the text content; the text position information is input into the embedding layer of the media content classification network, each dimension of the text position information is projected by embedding, and the projection results of each dimension are combined to obtain the embedding features of the text position information; the first image is subjected to feature extraction to obtain the features of the first image; next, the embedding features of the text content, the embedding features of the text position information and the features of the first image are connected together and input into the two-dimensional image classification layer for processing to obtain the media content classification to which the original image belongs.

[0044] Through the above processing, when the media content is classified, the text position information corresponding to the text content is fully considered, the features of the text content, the features of the text position information and the image features of the media content display area are all taken as the reference for the media content classification, so that the classification information of the media content can be judged by comprehensively considering the information in multiple dimensions, and the accuracy of the media content classification is improved. BRIEF DESCRIPTION OF DRAWINGS

[0045] Figure 1 FIG. 1 is a schematic diagram of the basic flow of the media content classification method in the application;

[0046] Figure 2 Fig. 1 is a schematic diagram of a basic structure of a media content classification network according to the present application;

[0047] Figure 3 Fig. 2 is a schematic diagram of a basic structure of a media content classification apparatus according to the present application;

[0048] Figure 4 Fig. 3 is a schematic diagram of a basic structure of an electronic device according to the present application. DETAILED DESCRIPTION

[0049] In order to make the purposes, technical means and advantages of the present application more clear, the present application is further described in detail below with reference to the accompanying drawings.

[0050] The basic idea of the present application is to combine the text semantics, text layout and image features of media content, and use neural networks to classify media content, thereby improving the accuracy of classification.

[0051] Figure 1 Fig. 4 is a schematic diagram of a basic flow of a media content classification method according to the present application. As shown in Fig. 4, the method comprises: Figure 1

[0052] Step 101, obtaining a first image of a media content display area and text content and corresponding text position information within the media content display area in an original image.

[0053] The original image is an image of a media content display board (e.g. an advertising board) taken. A part of the original image containing the media content display area is cut out and referred to as the first image. The process of obtaining the first image can be performed in existing ways, such as obtaining a region of interest (ROI) for matting, etc.

[0054] In the media content display area of the original image, the text content of the area and the corresponding text position information of the text content are obtained. The process of obtaining the text content and the corresponding text position information of the media content can be performed in various existing ways, such as sending the original image or the first image into a Chinese optical character recognition (OCR) engine to extract the text position and the text content.

[0055] In obtaining the text content and the corresponding text position information, the text content and the corresponding text position information can be obtained in units of text boxes. Specifically, in the OCR engine, the text of the media content display area can be divided into a plurality of text boxes, for each text box, the text content in the text box and the position information of the text box are obtained, and the position information and the text content corresponding to the same text box are associated.

[0056] ​The text position information can be information capable of locating the position of the text, which can be represented by coordinates or related information of the position of the text part. As a simple example, when the text content and the corresponding text position information are obtained in the unit of a text box, the text position information can be the coordinates of the specified boundary points of the corresponding text box (for example, the coordinates of the left top point and the right bottom point of the text box, or the coordinates of the left bottom point and the right top point of the text box), or can be the coordinates of the center point of the corresponding text box and the width and height information of the text box.

[0057] Step 102, input the first image, the text content and the text position information obtained in step 101 into the trained media content classification network.

[0058] The media content classification network is a pre-trained neural network, and the specific structure can be as shown in Figure 2 The specific details of various processes will be described in detail in the corresponding steps below.

[0059] When the text content is input into the media content classification network, it can be input in the unit of a text box. Assuming that the text content input length of the media content classification network is L text boxes, the text content obtained in step 101 is arranged in the specified text box order (for example, the text content corresponding to each text box is arranged in the order of the text box position from top to bottom and from left to right), and a sequence A with a length of L is formed. If the number of text boxes corresponding to the text content obtained in step 101 is greater than or equal to L, the text content of L text boxes is retained to form the sequence A, and the text content corresponding to the remaining text boxes is deleted. If the number of text boxes corresponding to the text content obtained in step 101 is less than L, the text content corresponding to all text boxes is supplemented, and the sequence length is supplemented to L. The supplemented content can be pre-set or randomly generated. The value of L is set according to the experience value of the number of text boxes of the media content display card in various scenarios.

[0060] When the text position information is input into the media content classification network, the same method as the text content is used.

[0061] Step 103, performing semantic feature extraction on the text content to obtain the semantic feature of the text content.

[0062] In this step, the semantic feature of the text character is obtained, and the specific processing can adopt the existing processing method. For example, the obtained Embedding feature can be used as the semantic feature by processing through the Embedding layer. The Embedding layer processing text character to obtain the Embedding feature can be performed in the existing manner, which will not be described here.

[0063] The semantic feature extraction processing can be generally performed in the unit of text box, and the semantic feature of the corresponding text content is obtained for each text box.

[0064] In step 104, the text position information is input into the Embedding layer of the media content classification network, each dimension of the text position information is projected by embedding, the projection results of the respective dimensions are combined, and the embedding feature of the text position information is obtained.

[0065] When the text position information is processed by the Embedding layer, each dimension of the text position information can be projected by embedding, the projection results of the respective dimensions are combined, and the embedding feature of the text position information is obtained. Specifically, taking the upper left and lower right point coordinates of the text box as the text position information as an example, the upper left and lower right point coordinates of the corresponding text box can be obtained first, that is, [x1, y1, x2, y2], then each coordinate is projected by embedding to obtain the corresponding embedding vector, that is, four vectors are obtained by four embedding projections, and finally the four vectors are added to obtain the embedding feature of the text position information corresponding to the corresponding text box.

[0066] For convenience of processing, preferably, after the coordinates are obtained, all the coordinates can be normalized to 0-1 according to the width and height of the image, and multiplied by 1000 and rounded, and finally the obtained coordinates are integers in the range of 0-1000. The processed coordinates are used for embedding projection and combination of the projection results.

[0067] As described above, the embedding feature of the text position information can be obtained, which reflects the feature information of the text position, that is, the text layout feature on the media content display card. The subsequent media content classification combined with the feature can improve the accuracy of media content classification (such as advertisement classification). Especially for the media content display card with partial occlusion, the classification accuracy can be effectively improved.

[0068] In step 105, the feature of the first image is extracted to obtain the feature of the first image.

[0069] In the present application, in addition to the semantic feature and layout feature of the text, the image feature of the media content display area is also extracted to utilize the visual image feature of the media content display card as the basis for media content classification.

[0070] The feature of the first image can be extracted in various existing image feature extraction methods, for example, the processing of the Embedding layer can also be used, which will not be described here.

[0071] The processes of the aforementioned steps 103, 104 and 105 are independent of each other, can be performed in parallel, or can also be performed in any order.

[0072] In step 106, the semantic features of the text content, the embedding features of the text position information and the features of the first image are connected together and input into a two-dimensional image classification layer for processing to obtain a media content classification to which the original image belongs.

[0073] This step performs fusion processing of the three features. The specific fusion processing can be performed as follows:

[0074] a. The semantic features of the text content and the embedding features of the text position information are connected together to obtain first features;

[0075] b. The features of the first image are transformed into two-dimensional features, and the size of one dimension is the same as that of the first features;

[0076] c. The first features and the transformed features of the first image are connected together.

[0077] The processing of a and b is independent of each other, can be performed in parallel, or can also be performed in any order.

[0078] Specifically, in step a, taking processing in units of text boxes as an example, it is assumed that the size of the semantic features is Lxd1, and the size of the embedding features of the text position information is Lxd2, L is the number of text boxes, the semantic features of the text content and the embedding features of the text position information corresponding to the same text box are spliced together, i.e. e_fuse = concat(e_text + e_coor), where e_text and e_coor represent the semantic features of the text content and the embedding features of the text position information respectively, concat represents splicing of matrices, to obtain fusion features e_fuse (i.e. first features), and the size is Lx(d1+d2).

[0079] In step b, the features of the first image are transformed into two-dimensional features. Generally, the image features are three-dimensional features, and for fusion with the first features e_fuse, it is necessary to convert the three-dimensional features of the image into two-dimensional features, and the size of the converted two-dimensional features is the same as that of one dimension of the first features e_fuse, for example, into a two-dimensional feature with a dimension of d1+d2. The three-dimensional features can be converted into two-dimensional features in various existing ways, for example, the size of the three-dimensional features can be first transformed through a linear layer of fc full connection, and then the three-dimensional features after size transformation are reshaped to realize dimension conversion, to obtain a two-dimensional feature map, and the size is assumed to be L1x(d1+d2).

[0080] Finally, in step c, the first feature e fuse obtained in step a and the converted two-dimensional feature obtained in step b are concatenated to obtain a final modal fusion feature f out.

[0081] The modal fusion feature f out is input into a classification layer, and classification processing is performed in a conventional manner to obtain the media content classification to which the original image belongs. The classification network can be various network structures capable of achieving classification, which will not be described here. The classification layer can first perform two-dimensional feature extraction on the modal fusion feature f out, and the feature extraction can be performed in the manner of two-dimensional image feature extraction, and then classification processing is performed.

[0082] At this point, the method flow of the present application Figure 1 is completed.

[0083] In the above processing, the first image, the text content and the text position information are processed by using a pre-trained media content classification network to obtain the semantic feature of the text, the layout feature of the text and the visual feature of the image as the basis for final classification. The media content classification network can be the neural network structure shown in the foregoing Figure 2 , and by pre-inputting standard training samples, the parameters of the neural network structure are trained, the classification loss function is calculated in the final classification layer, and the neural network parameters are adjusted according to the calculation result, so that a media content classification network capable of achieving accurate media content classification according to the foregoing processing is obtained. To ensure the classification accuracy of the media content classification network, the standard training sample is usually a media content display card image including text and various typical text layouts, and the text content and the text position information are extracted from the media content display card image and input into the media content classification network as training samples together with the media content display card image for training.

[0084] Through the above processing of the present application, the media content image, the media content layout and the semantic information of the media content are used to effectively improve the accuracy of media content classification (such as advertisement classification), and the problem of classification failure or reduced classification accuracy caused by media content occlusion can be effectively improved.

[0085] The present application also provides a media content classification device which can be used to implement the above-mentioned media content classification method. Figure 3 A schematic diagram of the basic structure of the media content classification device in the present application is shown in FIG. 1. As Figure 3 shown, the device includes an acquisition unit and a classification network unit.

[0086] The obtaining unit is configured to obtain a first image of a media content display area and text content and corresponding text position information in the media content display area in an original image;

[0087] The classification network unit is configured to process the first image, the text content and the text position information by using a pre-trained media content classification network to determine a media content category to which the original image belongs.

[0088] Specifically, the processing of the first image, the text content and the text position information includes:

[0089] performing semantic feature extraction on the text content to obtain semantic features of the text content;

[0090] inputting the text position information into an embedding layer of the media content classification network to perform embedding projection on each dimension of the text position information, combining projection results of the dimensions to obtain embedding features of the text position information;

[0091] performing feature extraction on the first image to obtain features of the first image;

[0092] connecting the semantic features of the text content, the embedding features of the text position information and the features of the first image together and inputting them into a two-dimensional image classification layer for processing to obtain the media content category to which the original image belongs.

[0093] Optionally, in the classification network unit, the processing of connecting the semantic features of the text content, the embedding features of the text position information and the features of the first image together can specifically include:

[0094] connecting the semantic features of the text content and the embedding features of the text position information together to obtain first features;

[0095] transforming the features of the first image into two-dimensional features with a dimension size same as that of the first features;

[0096] connecting the first features and the transformed features of the first image together.

[0097] Optionally, the text content and the corresponding text position information can be obtained in the obtaining unit in a text box unit.

[0098] In the classification network unit, the semantic feature extraction on the text content and the processing of the text position information by the embedding layer can both be performed in a text box unit.

[0099] Optionally, in the classification network, when connecting the semantic features of text content and the embedding features of text position information together, the semantic features of the text content and the embedding features of text position information corresponding to the same text box are connected together.

[0100] Optionally, the text position information can be: the coordinates of the specified boundary point of the corresponding text box, or the coordinates of the center point of the corresponding text box and the width and height information of the text box.

[0101] This application also provides a computer-readable storage medium that stores instructions, which, when executed by a processor, can perform the steps in the media content classification method described above. In practical applications, the computer-readable medium may be included in the devices / apparatus / systems of the above embodiments, or it may exist independently and not assembled into the device / apparatus / system. The computer-readable storage medium stores instructions, which, when executed by a processor, can perform the steps in the media content classification method described above.

[0102] According to the embodiments disclosed in this application, the computer-readable storage medium can be a non-volatile computer-readable storage medium, such as including but not limited to: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof, but not intended to limit the scope of protection of this application. In the embodiments disclosed in this application, the computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0103] Figure 4 An electronic device is also provided for this application. For example... Figure 4 As shown, it illustrates a structural schematic diagram of the electronic device involved in the embodiments of this application, specifically:

[0104] The electronic device may include a processor 401 with one or more processing cores, a memory 402 with one or more computer-readable storage media, and a computer program stored in the memory and executable on the processor. When the program in the memory 402 is executed, a media content classification method can be implemented.

[0105] Specifically, in practical applications, this electronic device may also include components such as a power supply 403 and an input / output unit 404. Those skilled in the art will understand that... Figure 4The structure of the electronic device shown in the figures does not constitute a limitation on the electronic device, and can include more or fewer components than shown, or combine some components, or arrange the components differently. Among them:

[0106] The processor 401 is the control center of the electronic device, connects the various parts of the entire electronic device through various interfaces and lines, performs various functions of the server and processes data by running or executing software programs and / or modules stored in the memory 402, and calling data stored in the memory 402, thereby overall controlling the electronic device.

[0107] The memory 402 can be used to store software programs and modules, that is, the above-mentioned computer readable storage medium. The processor 401 performs various function applications and data processing by running the software programs and modules stored in the memory 402. The memory 402 can mainly include a storage program area and a storage data area, wherein the storage program area can store an operating system, application programs required by at least one function, etc.; the storage data area can store data created according to the use of the server, etc. In addition, the memory 402 can include a high-speed random access memory, and can also include a non-volatile memory, such as at least one magnetic disk storage device, a flash memory device, or other volatile solid-state memory device. Accordingly, the memory 402 can also include a memory controller to provide the processor 401 with access to the memory 402.

[0108] The electronic device also includes a power supply 403 for powering various components, which can be logically connected to the processor 401 through a power management system, so as to realize functions such as management of charging, discharging, and power consumption management through the power management system. The power supply 403 can also include one or more direct current or alternating current power supplies, recharging systems, power failure detection circuits, power converters or inverters, power status indicators, and any other components.

[0109] The electronic device can also include an input / output unit 404, which can be used to receive input digital or character information, and generate keyboard, mouse, joystick, optical signal input related to user settings and function control. The input / output unit 404 can also be used to display information input by the user or information provided to the user and various graphical user interfaces, which can be composed of graphics, text, icons, video and any combination thereof.

[0110] The above only describes the preferred embodiments of the present application and does not limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

Claims

1. A method of classifying media content, characterized by, The method comprises the following steps: acquiring a first image of a media content display area and text content and corresponding text position information in the media content display area in an original image; processing the first image, the text content and the text position information by using a pre-trained media content classification network to determine a media content classification to which the original image belongs; wherein the processing of the first image, the text content and the text position information comprises: performing semantic feature extraction on the text content to obtain semantic features of the text content; inputting the text position information into an embedding layer of the media content classification network to perform embedding projection on each dimension of the text position information, combining the projection results of the respective dimensions to obtain embedding features of the text position information; performing feature extraction on the first image to obtain features of the first image; connecting the semantic features of the text content, the embedding features of the text position information and the features of the first image together and inputting them into a two-dimensional image classification layer for processing to obtain the media content classification to which the original image belongs.

2. The method of claim 1, wherein, The connecting of the semantic features of the text content, the embedding features of the text position information and the features of the first image together comprises: connecting the semantic features of the text content and the embedding features of the text position information together to obtain first features; transforming the features of the first image into two-dimensional features with the same dimension size as one dimension of the first features; and connecting the first features and the transformed features of the first image together.

3. The method of claim 1, wherein, The semantic feature extraction on the text content and the processing of the text position information by the embedding layer are both performed in units of text boxes.

4. The method of claim 3, wherein, When the semantic features of the text content and the embedding features of the text position information corresponding to the same text box are connected together.

5. The method of claim 1, 2, or 3, wherein, The text position information is coordinates of specified boundary points of a text box or coordinates of a center point of a text box and width and height information of the text box.

6. A media content classification apparatus characterized by comprising: The method comprises the following steps: an acquisition unit and a classification network unit; the acquisition unit is configured to acquire a first image of a media content display area and text content and corresponding text position information in the media content display area in an original image; the classification network unit is configured to process the first image, the text content and the text position information by using a pre-trained media content classification network to determine a media content classification to which the original image belongs; wherein in the classification network unit, the processing of the first image, the text content and the text position information comprises: performing semantic feature extraction on the text content to obtain semantic features of the text content; inputting the text position information into an embedding layer of the media content classification network, embedding projecting each dimension of the text position information, and combining projection results of the respective dimensions to obtain an embedding feature of the text position information; performing feature extraction on the first image to obtain a feature of the first image; concatenating the semantic feature of the text content, the embedding feature of the text position information, and the feature of the first image together, and inputting the concatenated result into a two-dimensional image classification layer for processing to obtain a media content classification to which the original image belongs.

7. The apparatus of claim 6, wherein, In the classification network unit, the concatenating the semantic feature of the text content, the embedding feature of the text position information, and the feature of the first image together comprises: concatenating the semantic feature of the text content and the embedding feature of the text position information together to obtain a first feature; transforming the feature of the first image into a two-dimensional feature with a dimension size same as that of the first feature; and concatenating the first feature and the transformed feature of the first image together.

8. The apparatus of claim 6, wherein, In the obtaining unit, the text content and the corresponding text position information are obtained in a text box unit. In the classification network unit, the semantic feature extraction on the text content and the processing of the embedding layer on the text position information are both performed in a text box unit.

9. The apparatus of claim 8, wherein, In the classification network, when the semantic feature of the text content and the embedding feature of the text position information are concatenated together, the semantic feature of the text content and the embedding feature of the text position information corresponding to the same text box are concatenated together.

10. The apparatus of claim 6, 7, or 8, wherein, The text position information comprises coordinates of a specified boundary point of a text box, or coordinates of a center point of a text box and width and height information of the text box.

11. A computer readable storage medium having stored thereon computer instructions, wherein, The instructions, when executed by a processor, can implement the media content classification method of any one of claims 1-5.

12. An electronic device, comprising: The electronic device at least includes a computer readable storage medium, and further includes a processor. The processor is configured to read executable instructions from the computer readable storage medium, and execute the instructions to implement the media content classification method of any one of claims 1-5. The electronic device at least includes a computer readable storage medium, and further includes a processor. The processor is configured to read executable instructions from the computer readable storage medium, and execute the instructions to implement the media content classification method of any one of claims 1-5.

Images